Man Not Machine: How False Memories Make Us Human

Art by Thao Do.

Where were you when you found out that the Twin Towers had been hit? You probably remember. Many people recall particularly shocking events in striking clarity, a phenomenon known as ‘flashbulb memory’. However, most memories are not like this: human memory is very unreliable, with certain things being remembered clearly and other aspects forgotten altogether (and in fact, even flashbulb memories may be more inaccurate than they seem). Even more worryingly, we ‘remember’ things that never actually happened. These false memories have profound implications for our understanding of human memory.

You might think of the brain as a computer, with memories being encoded and stored onto the hard drive, and retrieved when needed in the state they were saved. This tendency to compare the brain to the latest technology is not something new: in the past the brain has been compared to a water clock, a factory and a telephone exchange, when each of these were considered modern inventions.

However, as similar as they may first seem, the brain does not store memories like a computer stores information. In fact, the brain is much more unreliable. This is because computer memory is ‘reproductive’, so the memory is retrieved exactly as it was encoded. Human memory on the other hand is described as ‘reconstructive’, so the memory is built up from what we remember of what happened, fleshed out with our prior knowledge of the world to make a complete memory.

Not only does this mean that memories are incomplete, but it can also result in false memories being created. If an event is only vaguely remembered, it can be easy to falsely remember something that wasn’t there, but which makes sense in the context. For example, if a list of words related to ‘sleep’ (such as tired, nap, dream, doze, slumber) is presented to people to learn, they are likely to falsely recall that the word ‘sleep’ appeared in the list. That is, they have created a false memory that fits within the context of the situation.

Obviously, for daily life our memory is perfectly adequate, but in certain situations it is necessary to bear in mind the flaws of human memory. One instance in which the fallibility of human memory can be a problem is when eyewitnesses are asked to give their memories of events. In 1974, Loftus and Palmer demonstrated this by showing participants videos of two cars colliding, then asked them what speed they thought the cars were going. However, for different participants, they used different wording: they asked how fast the cars were going when they ‘contacted’, ‘hit’, ‘bumped’ or ‘smashed’ each other. You can see some of the clips in this video. The more vivid the word used to describe the impact, the faster participants’ estimated speed was (the difference was as much as 10mph!). And these altered memory effects persisted: the participants were asked a week later if there was any broken glass in the video (there wasn’t). Those who’d heard ‘smashed’ were far more likely to say ‘yes’ than those who had heard ‘hit’. This demonstrates the significant effect that lawyers’ leading questions could have on witnesses.

So each time we retrieve a memory, it can be changed. This means that the more often we remember something, the less accurate the memory may be, because it could be altered by the circumstances in which it was remembered. So our clearest memories (the ones we remember most) are possibly the least true.

But why do we have this reconstructive memory system if it’s so unreliable? It is a matter of capacity. We simply cannot remember every detail of everything we experience, so we remember some things more clearly than others. It is not as simple as pressing the save button and a document being encoded exactly in memory – if a word stands out in a list either perceptually (different colour), semantically (has a different meaning to other words in the list) or emotionally (has emotional significance whereas the other words don’t), it is more likely to be remembered. And it is useful to remember things that stand out because they are more likely to be important to survival.

Being able to reconstruct a full memory from incomplete scraps means that our memory system is very efficient, so we can remember vast amounts of information – especially that which is most important. But it does leave our memory vulnerable to manipulation and the creation of false memories. In this way, human memory is much less reliable than a computer, and for good reason. So perhaps if we want to create a computer that truly resembles a human brain, we should not aim for perfection, but for a few specifically human flaws.